Towards Patient-Specific Deformable Registration in Laparoscopic Surgery

Alberto Neri; Veronica Penza; Nazim Haouchine; Leonardo S. Mattos

cs.CVApr 14, 2026

Towards Patient-Specific Deformable Registration in Laparoscopic Surgery

Authors: Alberto Neri, Veronica Penza, Nazim Haouchine, Leonardo S. Mattos

Organizations: Biomedical Robotics Lab, Istituto Italiano di Tecnologia, Genoa, Italy · Department of Computer Science, Bioengineering, Robotics and Systems Engineering (DIBRIS), University of Genoa, Italy · Harvard Medical School, Brigham and Women’s Hospital, Boston, USA

Abstract

Unsafe surgical care is a critical health concern, often linked to limitations in surgeon experience, skills, and situational awareness. Integrating patient-specific 3D models into the surgical field can enhance visualization, provide real-time anatomical guidance, and reduce intraoperative complications. However, reliably registering these models in general surgery remains challenging due to mismatches between preoperative and intraoperative organ surfaces, such as deformations and noise. To overcome these challenges, we introduce the first patient-specific non-rigid point cloud registration method, which leverages a novel data generation strategy to optimize outcomes for individual patients. Our approach combines a Transformer encoder-decoder architecture with overlap estimation and a dedicated matching module to predict dense correspondences, followed by a physics-based algorithm for registration. Experimental results on both synthetic and real data demonstrate that our patient-specific method significantly outperforms traditional agnostic approaches, achieving 45% Matching Score with 92% Inlier Ratio on synthetic data, highlighting its potential to improve surgical care.

Figures & tables

Figure 1: The network uses KPConv to detect keypoints and extract their features, refines these features through self- and cross-attention, and then feeds them into the decoder and matching module to predict dense point correspondences.

Figure 2: Qualitative results of predicted matches. Blue points denote the preoperative point cloud, while orange points represent the partial and deformed intraoperative cloud. Red lines indicate incorrect matches, and green lines highlight exact matches relative to the ground truth.

Methods	MS (%)	IR (%)	Avg. #MP
FPFH	$0.03\pm 0.11$	$0.03\pm 0.11$	$<1$
Leopard	$1.6\pm 1.5$	$2.4\pm 2.2$	$4\pm 3.75$
LiverMatch	$13\pm 9$	$23\pm 13$	$39\pm 27$
Ours - ML + CL	$36\pm 8$	$90\pm 4$	$90\pm 20$
Ours - Full	$45\pm 7$	$92\pm 3$	$112.5\pm 17.5$

Table 1: Matching Metrics

Panel of Figure 3 — Figure 3: Non-rigid registration results on IRCAD-Liver1: Source point cloud is represented with gray points while target point clouds are represented in black wireframe. Green lines represent the matching correspondences and FEM models are represented with grid cells. Registered point clouds are presented at t=0, t=T/2 and t=T are shown in 4 different views.

Met.	Case0	Case1	Case2	Case3	Case4	Case8	Case10	Case5	Case6	Case7	Case9	Case11	Case12	Avg
HD	$11.12$	$8.44$	$6.62$	$6.23$	$15.45$	$10.17$	$12.04$	$8.12$	$5.65$	$6.93$	$6.88$	$5.98$	$7.75$	$8.45\pm 3.60$
FRE	$27.68$	$18.68$	$12.53$	$12.23$	$32.98$	$26.68$	$23.02$	$10.43$	$9.46$	$15.04$	$8.84$	$10.54$	$9.83$	$15.90\pm 6.41$

Table 2: Registration results on DePoLL dataset (results in mm).

Panel of Figure 4 — Figure 4: Qualitative results of non-registration on six cases from the DePoLL dataset are presented, showing the endoscopic image, the matching, and the registration from left to right. Preoperative point clouds are represented in gray, intraoperative point clouds in black, and correspondences are indicated with green lines.

Laparoscopic Surgery

See all 21 papers

Preoperative-to-intraoperative Liver Registration for Laparoscopic Surgery via Latent-Grounded Correspondence Constraints
Mar 2, 2026Ruize Cui, Jialun Pei, Haiqiao Wang +42D-3D Image Registration Laparoscopic Surgery
The Hong Kong Polytechnic University, Hong Kong, China · The Chinese University of Hong Kong, Hong Kong, China
Test Time Adaptation Methods for Point Cloud Registration in Laparoscopic Surgery
Aug 3, 2026Nina Bodelot, Soufiane Belharbi, Eric GrangerSynthetic-to-Real Domain Adaptation Test-Time Adaptation
LIVIA, Dept. of Systems Engineering, ETS Montreal, Canada
Depth Augmented and FE Free 3D/2D Liver Registration for Laparoscopic Liver AR
Feb 19, 2026Hanyuan Zhang, Lucas He, Runlong He +62D-3D Image Registration AR in Surgery
UCL Hawkes Institute, University College London, London WC1E 6BT, UK · Unit for Lifelong Health and Ageing at UCL, University College London, London WC1E 7HB, UK · Medtronic plc., London, UK +1

Mar 2, 2026J/K move · Enter open · S save

Preoperative-to-intraoperative Liver Registration for Laparoscopic Surgery via Latent-Grounded Correspondence Constraints

Ruize Cui, Jialun Pei, Haiqiao Wang, Jun Zhou+3

The Hong Kong Polytechnic University, Hong Kong, China · The Chinese University of Hong Kong, Hong Kong, China

2D-3D Image Registration Laparoscopic Surgery Medical Image Registration Cross-Modal Image Registration Deformable Image Registration

Open paper AskPDF arXiv

In laparoscopic liver surgery, augmented reality technology enhances intraoperative anatomical guidance by overlaying 3D liver models from preoperative CT/MRI onto laparoscopic 2D views. However, existing registration methods lack explicit modeling of reliable 2D-3D geometric correspondences supported by latent evidence, leading to limited interpretability and potentially unstable alignment in clinical scenarios. In this work, we introduce Land-Reg, a correspondence-driven deformable registration framework that explicitly learns latent-grounded 2D-3D landmark correspondences as an interpretable intermediate representation to bridge cross-modal alignment. For rigid registration, Land-Reg embraces a Cross-modal Latent Alignment module to map multi-modal features into a unified latent space. Further, an Uncertainty-enhanced Overlap Landmark Detector with similarity matching is proposed to robustly estimate explicit 2D-3D landmark correspondences. For non-rigid registration, we design a novel shape-constrained supervision strategy that anchors shape deformation to matched landmarks through reprojection consistency and incorporates local-isometric regularization to alleviate inherent 2D-3D depth ambiguity, while a rendered-mask alignment enforces global shape consistency. Experimental results on the P2ILF dataset demonstrate the superiority of our method on both rigid pose estimation and non-rigid deformation. Our code will be available at https://github.com/cuiruize/Land-Reg.

Abstract

Laparoscopic Surgery

Explore similar work

Preoperative-to-intraoperative Liver Registration for Laparoscopic Surgery via Latent-Grounded Correspondence Constraints

Test Time Adaptation Methods for Point Cloud Registration in Laparoscopic Surgery

Depth Augmented and FE Free 3D/2D Liver Registration for Laparoscopic Liver AR